[00:00] Lexi: This week, LinkedIn published 3 years of research saying that almost every B2B buying group is now going to ChatGPT before they'll get on a call with your sales team. And that the AI isn't really discovering anybody. It's retrieving whoever the internet already trusts.
[00:14] Lexi: Then a company called DISQO turned around and said they can finally measure whether your advertising moves any of it.
[00:20] Lexi: One of those stories my 2 builders basically shrugged at. The other one genuinely split them.
[00:27] Lexi: Hi, and welcome back to the Aloudable HQ. This is Assistant to the CMO, I'm your host, Lexie Meskouris, and the idea here is pretty simple. Every week we take whatever just happened in AI and marketing, and we actually pull it apart.
[00:40] Lexi: Now, I don't build any of this myself. So each week I sit down with 2 people who do, Will Nash and Harjot Singh. Between them it's, honestly, over 20 years of building tech and brands from the ground up. And the best part is they've got no horse in the race. They don't sell the things we cover. They just see them differently.
[00:56] Lexi: And this week it's about who the AI decides to recommend, and whether you can buy your way into that. LinkedIn's got a 3 year study out, Mimi Turner's team, with Bain and half the industry attached to it. Their headline number is that 94% of B2B buying groups now use an LLM before they'll speak to sales. And their big claim is that AI is a retrieval system, not a discovery engine.
[01:16] Lexi: Will, I'm starting with you, because I have a feeling you've already got a pinch of salt ready. Is this a real shift, or is it a rebrand?
[01:23] Will: Good question. So, um, I think the honest answer is a bit of both, but if I had to come down on one side, I'd say it's more of a continuation than a shift. Because, I mean, the underlying thing they're pointing at, which is that people are using these tools as part of how they buy, that's, that's clearly a real signal, and it has been for, you know, at least a year now. So that bit's true. What I'd, what I'd sort of push back on is the framing that this is some, some entirely new phenomena, because, um, it isn't really. It's a continuation of a trend that's been running for, for a good while.
[01:53] Will: And then, and then the other thing, which I think you have to just, you have to hold in your head when you read this, is that, um, you kind of have to take LinkedIn's research with a pinch of salt. And I don't, I don't want to be too cynical about it, but the reason I say that is, is because their whole model is selling reputationally-based advertising, right, versus the more traditional discovery-based advertising. So when they tell you it's all shifting towards retrieval and reputation, they are, they're quite heavily incentivised to see it that way. That's, that's just worth sitting with before you take the 94% and run.
[02:18] Lexi: And that's the part I keep chewing on, because 3 years of research is still 3 years of research. Harjot, when you read that number, does it land as new information for you, or is Will right that we've all been living this for a while?
[02:30] Harjot: Yeah, no, I mean, I'm with Will on this. To be honest with you, when I read that number, um, it just seemed kind of obvious? Like, all it really shows is how widespread AI use is now, right, it's, it's not anything brand new. It's, it's really just surfacing what we all already know about this. And, and I think that's the thing, like, LinkedIn published it. So, honestly, yeah, it just feels like a piece to keep things going for them.
[02:51] Harjot: But, um, the bit that's actually interesting to me, and this is where I sort of, I keep landing, is, is why people are doing it at all. Because, you know, if you think about it, this is no different to when you go and ask 10, 15, 20 people, as many people as you possibly can, just so that you feel more confident before you commit to something. It's, it's a comfort thing, more than anything.
[03:09] Harjot: And what's kind of wild, um, is that people trust the AI a lot more than they trust other people, and, and often even more than they trust themselves, right? So they end up sort of outsourcing the judgment to it. Though, um, to be fair, there's a version of this that's completely fine, right? Like, as long as they're not using it to actually make the judgment, but they're using it to aggregate everything so that they can make the judgment themselves, then that's, that's pretty cool. That's a very sensible thing to do. But, um, yeah, to be honest with you, when you look at why they're reaching for it in the first place, it's a really emotional thing. It's about security, it's about feeling safe in the decision, you know, rather than it being about anything, anything real.
[03:40] Lexi: Okay, but "anything real" is doing a lot of work in that sentence. If I'm a marketing leader and my buyer feels safe picking me, I'm not sure I care whether the feeling was real. I got picked.
[03:50] Will: I think Harjot's right on the emotional thing, genuinely, like the comfort, the wanting to feel safe in the decision, that, that is real and I wouldn't argue with it. But, um, I'd, I'd gently push on the "rather than anything real" bit, because there is actually something real underneath the feeling, and I think that's, that's the part that's worth pulling apart.
[04:08] Will: So, so what these models are doing, certainly the ones we've had for, call it the last 6 to 9 months, is, it's kind of a mixture of things. They're leaning on the training data, the pre-training data, and then there's, there's specific post-training on top of that, right, that's, that's the bit everyone sort of knows about. But the newer ones are increasingly doing 2 other things. They're doing chain-of-reasoning, and, and crucially they're performing live web searches while they answer you. So they can now, they can go and look outside of their own training data.
[04:31] Will: And, um, and I think that's the thing that's genuinely changed, because, Harjot, to your point, the feeling might be totally emotional, but what's feeding the feeling is a, is a retrieval layer, and that layer is newly, sort of, contestable. Where before, you know, when these models first came out, whoever was around and, and well-represented at the moment all the training data got hoovered up, they had this, this big incumbent advantage baked in. And what the web search does is it starts to, to erode that a bit.
[04:58] Lexi: Hold on, I need to stop you, because you've said retrieval layer twice now and I want to make sure everybody's still with us. What is that, actually? Say it like I've never heard the phrase.
[05:07] Will: Yeah, sure, and it's, it's a fair pull-up, because I do, I say it like everyone knows what I mean. So, um, the way I'd, the way I'd explain it to a customer is, when you ask one of these things a question, it goes off and, and it looks things up in the moment, right? And the retrieval layer is basically just, it's the pool of material it's able to go and pull when it goes looking. So that's, that's your own website, and it's the, the third-party sites that it happens to trust. That's it, really. It's, it's what's reachable and pullable at the moment it's trying to answer a question about you.
[05:33] Will: And, and I think the reason that matters, the sort of practical, builder's version of it, is, if you're, if you're the challenger in your category, that's, that's the thing you actually go and attack. So, um, the first thing is just making sure your website can be easily taken in by these models. So, so there's a file, your llms.txt, that's, that's part of it. And then FAQs, Q&As, just, just all of the material that's sat there ready for when the model shows up at your site, or, or honestly when it shows up at some third-party site trying to find you, so that there's, there's actually something clean for it to pull.
[06:04] Will: Because if there's nothing there, it, it just can't retrieve you well. And then the other bit, which, which people find slightly annoying but it's, it's true, is that the models do over-weight certain websites. So Reddit's the, that's the obvious one everybody knows, but there's, there's a whole bunch of others. So part of the game is just, is being present and, and accurately represented in the places the model already leans on.
[06:24] Lexi: So that's the challenger. But a lot of the people listening to this are not the challenger. They're the big incumbent with 20 years of reputation and a brand everybody already knows. Does any of this even touch them?
[06:36] Will: So it does, yeah, but, but in a, in a much less dramatic way, I think. So if you're, if you're that longstanding business, 20 years of reputation, the models already know who you are, they're already very aware of you, then honestly, um, this is, this is just a continuation of the pressure that CMOs have been under for years anyway. Which is this, this sort of slow shift towards earned reputation and, and away from pure discoverability. And that, that just keeps going. So, so you lean on the trust you've already got, that's, that's real and it's a genuine advantage, but, but the point is you can't just sit on it. You've got to keep, keep building on it.
[07:06] Will: And then, um, the bit that I actually think is a little bit underrated, is, if you're an incumbent and you're one of the, the rare people who can genuinely think in, in 5 and 10 year cycles, and I know that's, that's not most people, but if you can, it's, it's worth starting to think now about how information about you gets taken into the models of, of the future. How you're explained, how you're understood, what the model thinks you actually do.
[07:29] Will: And, and especially, this is the one, especially if you're going through any kind of M&A, or a rebrand, you really want that to be extremely clear and, and exposed on the retrieval layer. Because otherwise the model just, it gets confused about who you are now versus who you were, and, and it ends up recommending you for, for the wrong thing. So even for the incumbent, there's, there's a bit of housekeeping there that's worth doing.
[07:48] Lexi: But the thing I want to know is whether the small brand actually has a shot here, or whether we're just describing a nicer-sounding version of the big guys winning again. Because you can't manufacture 20 years of customer proof.
[08:00] Harjot: I think you've genuinely got a shot, yeah. Like, sure, you may not have 20 years of reviews, um, but maybe you make up for that with tons of articles, tons of content, all the, you know, the AEO-inducing material, so that your presence is as holistic as you can possibly make it. Because lack of presence, that's always a problem, and the AI absolutely picks up on it.
[08:18] Harjot: But, um, the thing is, you now have a way to actually compete with the big guys, right, because you can look at answer engine optimisation and, and find the things they can't compete on. So it's, it's David versus Goliath, to be honest with you. It's just, where are the areas you can actually win?
[08:32] Lexi: So if the AI is retrieving whoever's already trusted, the obvious next question is whether you can pay your way into that. And a company called DISQO says they've built the thing that finally tells you. They announced it on the 17th of June, they're calling it AI Search Lift, and they're calling it an industry first. It's a panel of consumers who've consented to be tracked, they run an exposed group against a control group, and they say they can measure whether your advertising moves your brand inside AI search. They ran it across 5 categories, automotive, insurance, beauty, CPG, travel. And none of it's independently verified yet. Will, from the engine room. Real measurement, or a proxy in a nice suit?
[09:10] Will: So, um, let me, let me give them the, the strong version first, because there is, there's a real thing here. There genuinely is a blind spot, right? Nobody has proof of whether there's a connection between paid media, the advertising you're running, and, and people then going into these LLMs. That, that gap is real, and, and it does matter, and, and to be fair to DISQO, they're at least pointing at it. So, so I'll give them that.
[09:34] Will: But, um, but then when I actually look at what it does, I, I come off the fence the, the other way, honestly. Because as far as I can see, at the moment this is really just a kind of, a replication of branded search. Like, all you're, all you're actually testing is, are people willing to see an advert and then go and search for you. That's, that's the thing being measured. It's just that in this case the search isn't happening on Google, the search is, is happening in an LLM. And, and so it's probably, it's probably not quite as impressive as, as they're saying it is.
[10:02] Will: And, and here's, here's the bit that I think matters most. What a CMO would actually want to know, the thing they'd, they'd really love an answer to, is, is there a link between the paid advertising and the, the recommendations that come out of the chatbot. Right? That, that's the golden question. And, and that is not what this does. This is, this is checking whether there's some uptick in the search. Which, which is a different thing. It's, it's related, but it's, it's a different thing.
[10:25] Lexi: You're doing the face you do when you don't agree, Harjot!
[10:28] Harjot: See, this is where I actually come off the fence the other way from Will. Because, look, Will, realistically only time will tell, right, but I think there's genuinely something quite interesting here, something a bit different that, that no one else is really doing. So I, I do rate it. I do respect it, for sure.
[10:41] Harjot: And, um, let me try and say why, because I think, I think there's a real thing here that's easy to miss. So, with the GEO and the AEO platforms, right, the way they work is, you make a change, and then some period later you can see the improvement come back. So you can measure, you know, how effective your optimisation actually was over time. That bit's fine.
[10:59] Harjot: But, um, the thing is, you have no insight into the bit in the middle. And the bit in the middle, that's the human. That's the box in the middle, right? You can see the input, the change you made, and you can see the outcome, the result that comes back, but the person sitting in between, how they actually thought about it, what happened in their head between your input and that outcome, you, you never see that. It's just, it's completely opaque to you. And that's the thing I think DISQO might actually be shining a light on, right, the box in the middle.
[11:22] Harjot: Now, um, to be honest with you, I'm, I'm still trying to wrap my head around whether this actually makes a big difference in practice. Like, I'm not fully there yet. But that's, that's the reason I rate it, and I don't think it's just branded search relabelled.
[11:35] Will: It's a good push, Harjot, genuinely, and I, I'll actually give you the box-in-the-middle thing, because I think that's, that's a fair description of the gap. The, the person in between the input and the outcome, you're right, that, that's genuinely opaque, and, and if something's shining even a bit of light in there, then, then fair enough, that's, that's worth taking seriously.
[11:52] Will: So, so let me, let me try and say where I'd be wrong, because I think that's, that's the honest way to do this. Um, where I'd be wrong is, is if you're a large business, and you run this kind of brand lift on, on traditional Google search, the normal thing, and then you run the same thing on an LLM, and, and you see a significant difference between the 2. Right? That, that's the scenario.
[12:15] Will: And, and I'll be honest, I don't, I don't really see why that would be true. But, but if it were, if it turned out that people were much more willing to go and research something in an LLM after they'd seen your advert than they were on Google, then, then yeah, that, that's a genuinely interesting insight, and, and you could absolutely turn that into something useful.
[12:33] Will: So I suppose the, the thing I'd actually want to see, the thing that would move me, is the delta. It's, it's your search lift versus your AI search lift, side by side. Because if, if those 2 just track each other, then, then to me it's just branded search, it's, it's the same behaviour, it's, it's just happening in an LLM instead of on Google. But if there's, if there's a real gap between them, then, then Harjot, you're onto something and I'd, I'd happily be wrong. That, that's the number I'd go looking for.
[12:57] Lexi: Okay, but somebody should run that experiment is not the same as somebody will pay for it. So who is this actually for? Give me the person with the budget.
[13:05] Harjot: And, and to your point, Will, the delta is exactly the thing, right, that's, that's the thing worth watching. But, um, on the who is it for, Lexie, I think I've actually got a specific person in mind. So this might be very helpful for a particular kind of company, and it's the ones doing performance-based marketing, right, where basically everything is based on data. The, the data-first teams.
[13:24] Harjot: Because here's the thing I see break for those teams. For them, the AEO work is very, very hard to justify. It's just, it's not as easy to justify as, you know, traditional SEO or attribution, because you don't have the same level of information there, right? You can't go to them and show them the number that they're used to seeing. It's, it's kind of invisible to the way they make decisions.
[13:41] Harjot: So, um, something like this could be a super, super fascinating offer for exactly those people, because it might, it might actually let them justify investing more into AEO in the first place. Like, it gives them a thing they can point at. So, yeah, perhaps it genuinely is an unlock, you know, for these larger companies that are doing performance-based marketing. That's, that's the sort of person I'd point it at, to be honest with you.
[14:02] Lexi: All right, so say that buyer is in the room and the contract's on the table. What's the one thing you'd make them check before they sign it? Will, you go first.
[14:10] Will: Yeah, for me it's, it's panel size, and, and specifically because it's the bit they haven't actually released yet. So, so I'd want to know, before I sign anything, what, what actually is the sample size here? Like, how many people are we, are we really talking about?
[14:23] Will: Because, um, if this turns out to be, you know, 2000 people, and, and worse, if it's, if it's 2000 men under 35, right, then that's a, that's a really, really small and, and narrow group. And, and you should be even more sceptical about what you're being shown off the back of it. Because at the moment we're, we're all sort of trusting a bit of a black box here, and, and the panel is, is the thing that either makes it credible or, or doesn't. So that, that's my one, really. I'd, I'd just want to see that panel number before I got at all excited about a lift they're showing me.
[14:54] Lexi: And you'd want something else entirely, Harjot.
[14:57] Harjot: Mine's different, yeah. Mine's, um, mine's really about methodology and trust. So, the thing is, being first to market with something like this, right, it doesn't actually mean it solves the problem. Those are, those are 2 completely different things, you know.
[15:10] Harjot: And, um, here's the thing, it's, it's not super clear who exactly this is for straight out of the gate. And I actually think that's kind of deliberate, because the moment they tell you precisely who it's targeted at, that becomes a claim, right, and then that's a claim they'd also have to go and support. So instead they've sort of said, hey, look, we've got this thing, it's a novel approach, come and have a look.
[15:27] Harjot: And, you know, my guess is they haven't gone too deep publicly into exactly how it solves the problem, because, to be honest with you, everyone's so early in all of this.
[15:36] Harjot: But here's, here's the real demand for me. If that panel is completely removed from how people actually think, right, then they've got a selection problem, plain and simple. So whatever the selection criteria is, and, and the methodology behind how the panel itself is built, that needs to be really well publicised, well socialised, so that there's genuine trust in it. Because ultimately, um, this whole thing is about trust. That's, that's what it comes down to.
[15:56] Harjot: So they need to publish, you know, what their pinnacle of trust is. Like, what is the thing that makes us go, okay, we can actually rely on them and rely on their panels. And then the last bit, right, is just, does it actually solve my problem? If I've got a data visibility problem, is this thing genuinely solving that, or, or is it solving something adjacent? So just, you know, double-check it's actually doing the thing.
[16:13] Lexi: That's the checklist for the person buying it. But plenty of people listening are running these things themselves, to do their own research. Does any of that change for them?
[16:23] Harjot: Yeah, it does, and it's sort of the flip side, right, because I've been talking about it as the CMO trying to reach these people. But if you're the CMO yourself, and you're the one using this kind of approach to go and do your research, then the thing you want to do is really test the system you've built. Your, your LLM, your ChatGPT-type sorter. Benchmark it against something you've actually done, or, or would do in real life, so that you've got a clear read on the difference. Otherwise, um, you don't really know what it's giving you.
[16:46] Lexi: So nobody's walking away from it, and nobody's moving money towards it either. If I take that to my CFO, it sounds like I'm asking to spend budget on a maybe.
[16:55] Will: And, and I think that's, that's exactly right, and I'd, I'd just try and make it a bit more precise, which is, treat this as a hypothesis. That's, that's the word I'd use. It's, it's a hypothesis, not a, not a finding yet.
[17:06] Will: So, so running the audit, honestly, great idea, especially if, if you've got the budget to go and do it. But, um, if you run it, and you find that, that some particular cohort starts searching once they've, they've been exposed to your paid media, that is, that is not the moment to go and start shifting your budget around in any significant way.
[17:21] Will: Because there's, there's a few steps in between. You'd want to check whether it replicates over Google, which is, you know, that's the delta again. And, and really the bar, the thing that would actually justify moving money, is, is a significant delta, sustained over a, a significant period, with a panel that you, that you actually trust. That's, that's the whole package. And until, until you've got that, I think, I think the honest verdict is, it's, it's maybe time to explore, but it's, it's not time to start putting your money where your mouth is yet.
[17:47] Lexi: Okay, last bit, and it's the part everybody actually shows up for. One thing a marketing leader could genuinely do on Monday. No budget, no re-platforming, done before lunch. Will, yours first, and keep it small.
[17:59] Will: Yeah, okay, keeping it small. So, um, the thing for me is, is that both of these stories, the, the buyability one and the search-lift one, they, they actually come back to the same place, which is, the, the trusted, over-indexed sources, they feed both things. They feed the training and they feed the retrieval. So, so that's, that's where I'd point somebody.
[18:20] Will: Which is, go and make sure there are, there are consistent, positive, accurate explanations of what you actually do sitting on those sites. Reddit's the obvious one, but, but also YouTube. And just, just be diligent about it, that, that the messages are well explained, that they, they build trust, that they don't misdirect anybody, and, and honestly that when they go out of date, somebody, somebody actually goes and changes them. And, and this isn't, this isn't a big campaign, right, lots of brands have their own Reddit accounts now, it's, it's totally normal. There's, there's a lot of free juice just kind of hanging around on those platforms.
[18:50] Will: But if you want the, the actual Monday move, the concrete one, um, it's this. Almost all of these chats, the way they're, they're formatted, is a Q&A. The user's sitting there asking, how does this work, can you do X, will you provide Y, are they, are they any good at this, are there, you know, bad recommendations floating around about them. It's, it's a very clear question-and-answer process.
[19:10] Will: So the, the source material you're feeding in, it, it has to mirror that format straight back. So go and think about, what's the first question you get from people. What's the, the big issue that comes up in the late-stage sales calls. Can you do X. Do you, do you have your cybersecurity accreditation. Whatever it, whatever it actually is. And then you put that on your website, clearly, front and centre, and you, you answer it. And the, the kicker on that is, even if it never makes it into the training data, it, it still goes into the retrieval data.
[19:39] Lexi: And you'd add the discipline to that, Harjot? Will's telling them to go and put the answers out there. What's the bit he's leaving out?
[19:45] Harjot: So, um, this is the thing I keep coming back to. These tools, right, they're only actually useful once you're using them over a period of time. Like, you can't just check it one day, one month, and go, yep, brilliant, we're done, we're looking good. That, that doesn't tell you anything. You have to trend it. That's, that's the first thing. You're watching the shape of it over time.
[20:00] Harjot: And then the second thing is, you want to correlate it against everything else, right? Your platforms, your measurements, all the other indicators you've already got. You correlate them and you check that they're all sort of pointing in the same direction. Because, um, ultimately, you know, all of this is influencing your bottom line one way or another. That's, that's what it's all about at the end of the day.
[20:17] Harjot: So here's the bit that matters. If this approach is suddenly showing you these amazing results, right, and it's just not propagating through the business, it's not touching the bottom line, it's not correlating with any of your other metrics, then, um, that's cause for concern. That's a thing to go and investigate.
[20:31] Harjot: And, and to be fair about it, it means one of 2 things. Either the tool is genuinely seeing something unique, something that none of your other indicators can see, and, you know, that would be super, super cool, and it could be the case, of course it could. It's just, it's not that common when you've already got lots of indicators, right? Or, um, and this is the more likely one, the tool is basically reporting vanity metrics, vanity results, and it's just not actually changing anything day to day.
[20:54] Harjot: So, yeah, the thing I'd leave everyone with, and honestly this is the whole thing for me, never skip baselining where you are right now and knowing what goals you actually want to achieve. Because if you don't have those 2 things, um, this is absolutely useless to you.
[21:07] Lexi: And that's the thing I keep coming back to. Both of these stories arrive dressed up as something brand new, and underneath, both of them are asking you the same unglamorous question, which is whether you actually know what's being said about you and whether you'd notice if it changed.
[21:23] Lexi: I loved this conversation. Honestly, I could have kept going on the DISQO stuff all day. Thank you both.
[21:29] Lexi: And that takes us to the end of this week's episode.
[21:32] Lexi: And if you're listening, thank you so much for being here. We'll be back every week with something new.
[21:37] Lexi: Before we go, I have to tell you something. Every voice you've heard today, all 3 of us, me included, it's all AI. But the opinions are absolutely ours, and the judgment behind every word is human. That's the part you can't replace. And that's what Aloudable does. We take what you'd write, turn it into a show in your own voice, and we keep the taste where it belongs, with a person.
[21:57] Lexi: And we're not selling you a lift number. We just make the show. So if you want to hear what that would sound like for your brand, there's a link in the show notes.
[22:05] Lexi: And that's the tea. I'm Lexie Meskouris, this has been Assistant to the CMO, and we will see you next week.